Privacy Enforcement Strategies in Discrete Event Systems via Observation Modification and Supervisory Control
Wei Ran Duan, Christoforos N. Hadjicostis, Zhiwu Li · 2021 European Control Conference (ECC) · 2021
We consider the problem of privacy enforcement in discrete event systems modeled with deterministic finite automata under partial observation. The notion of privacy considered in this paper is related to a set of secret events, whose occurrence needs to be indefinitely kept hidden from an external observer. We assume that the secret events are unobservable but (at least under some executions) revealable (inferable), in the sense that their occurrences can be indirectly determined with certainty after a finite number of subsequent observations. Our work proposes the use of alter functions (that can change the output observable events of a system) as a defense mechanism for preventing the detection of secret events. The alter function has the capability to replace the observed event with a fake (constructed) event or insert a fake event before the observed event. Supervisory control is used to determine the alterations that ensure that privacy can be enforced not only currently but also during subsequent operation of the system.